A species-discriminatory single-nucleotide polymorphism set reveals maintenance of species integrity in hybridizing European white oaks (Quercus spp.) despite high levels of admixture

Abstract

Background and aims: Hybridization and introgression play an important role in the evolution and diversification of plants. To assess the degree of past and current hybridization, the level of genetic admixture in populations needs to be investigated. Ongoing hybridization and blurred species separation have made it challenging to assign European white oak taxa based on leaf morphology and/or genetic markers and to assess the level of admixture. Therefore, there is a need for powerful markers that differentiate between taxa. Here, we established a condensed set of single-nucleotide polymorphism (SNP) markers to reliably differentiate between the three most common oak species in temperate European forests (Quercus robur, Q. petraea, Q. pubescens) and to assess the degree of admixture in a large set of selected Swiss populations.

Methods: A training set of 194 presumably pure reference samples from Switzerland and Europe was used to assign 633 test individuals with two different approaches (population genetic-based/Bayesian vs. assumption-free/discriminative classifier) using 58 selected SNPs from coding regions. Admixture was calculated at the individual and population level with the Shannon diversity index based on individual assignment probabilities.

Key results: Depending on the approach, 97.5-100 % of training individuals were assigned correctly, and additional analyses showed that the established SNP set could be further reduced while maintaining its discriminatory power. The two assignment approaches showed high overlap (99 %) in assigning training individuals and slightly less overlap in test individuals (84 %). Levels of admixture varied widely among populations. Mixed stands of Q. petraea and Q. pubescens revealed much higher degrees of admixture than mixed stands of the other two taxon pairs, accentuating high levels of gene flow between these two taxa in Switzerland.

Conclusions: Our set of SNPs warrants reliable taxon discrimination with great potential for further applications. We show that the three European white oak taxa have largely retained their species integrity in Switzerland despite high levels of admixture.

Keywords: Quercus; Admixture; SNPs; hybridization; taxon assignment; white oaks.

Fig. 1.

Genetic properties of the training set comprising 194 pure individuals of Quercus spp. from Switzerland and Europe (red circles) based on genotype data from 58 single-nucleotide polymorphisms. (A) Plot of the first two principal components (PC1 and PC2) of the principal component analysis; the variation explained by each PC is given in parentheses. (B) Unrooted neighbour-joining tree based on allelic differences.

Fig. 2.

Overview and comparison of three assignment approaches based on genotype data from 58 single-nucleotide polymorphisms. Each bar represents a single Quercus spp. individual and colours reflect assignment probabilities to respective clusters. Individuals (n = 827) are arranged according to training and test set. (A) Bar plot of Q values for STRUCTURE (Pritchard et al., 2000) without USEPOPINFO (K = 3). (C) Bar plot of assignment probabilities for STRUCTURE with USEPOPINFO (this model assumes that the predefined species clusters are correct). (E) Bar plot of support vector machine (SVM) assignment probabilities. (B) Correlations between assignment probabilities from STRUCTURE with USEPOPINFO (x-axis) and STRUCTURE without USEPOPINFO (K = 3, y-axis), respectively. (D) Correlations between assignment probabilities from STRUCTURE with USEPOPINFO (x-axis) and SVM assignment (y-axis), respectively.

Fig. 3.

Distribution of 58 locus-specific estimates of genetic differentiation [F_ST, based on pure individuals from the STRUCTURE (Pritchard et al., 2000) assignment with USEPOPINFO] between (A) Quercus petraea–Q. pubescens, (B) Q. robur–Q. petraea and (C) Q. robur–Q. pubescens.

Fig. 4.

Evaluation of the minimum number of single-nucleotide polymorphisms (SNPs) used for accurately assigning the training set with support vector machine. Accuracy denotes the proportion of correctly assigned Quercus spp. training individuals (>0.9 assignment probability). The input order of SNPs is ranked by highest F_ST (solid lines) or random (dashed lines, 10 000 iterations).

Fig. 5.

Principal component analyses of pure individuals of Quercus spp. using different underlying data types for taxon assignment. The plots show the first two principal components (PC1, PC2) and their explained variation in parentheses. (A) Genotype data from 58 single-nucleotide polymorphisms (SNPs) and (B) genotype data from eight microsatellites (nSSRs).

Fig. 6.

Admixture levels in populations of Quercus spp. (A) STRUCTURE (Pritchard et al., 2000) assignment with USEPOPINFO (K = 3, this model assumes that the predefined species clusters are correct) of 31 populations and (C) pairwise STRUCTURE assignment with USEPOPINFO (K = 2) of populations with a minimum of ten individuals after excluding individuals potentially belonging to the third species cluster. Each bar represents a single individual and colours reflect assignment probabilities. Per population, the averaged admixture index based on Shannon’s diversity (S) is indicated in red. Populations are ordered from high (left) to low (right) admixture index. The first digit of the population number reflects regions of origin in Switzerland: 1 = Jura; 2 = Swiss Plateau; 3 = Pre-Alps and Alps; 4 = Valais; 5 = Ticino; 6 = Grisons. (B) Histogram showing the frequency of S of all populations and (D) histogram for pairwise S based on results from (C).